The future of Dgraph is open, serverless, and AI-ready

At Hypermode, we believe that models will fundamentally change how apps are built over the next decade. Whether it’s the latest developments with generative AI or traditional machine learning approaches, incorporating models into an app allows for unmatched personalization, discovery, and productivity.

Building a model-native app includes, but can be much more than prompting a language model and returning the results directly to the user. It often involves enriching the prompt or results with an organization’s proprietary data. Or, it could be bringing natural language search indexes backed by embedding models directly into the app’s database.

No matter what way you slice it, harnessing and integrating data is critical for model-native apps. We were excited to acquire Dgraph Labs last year, in part because of the increased role we see knowledge graphs playing in incorporating specific, timely knowledge with AI models.

Graphs are the most natural structure for both humans and language models to understand relationships in data.

Beyond the release of native vector support in Dgraph v24, we’ve been working with a number of organizations adopting Dgraph as the knowledge graph foundation for their AI strategy. We’ve also seen teams discovering how graph databases can efficiently power app features like leaderboards, activity feeds, and network maps. While there’s energy around new AI-powered possibilities, the upcoming improved performance optimizations and query expressiveness benefit all types of use cases.

It’s encouraging what our users are building, but also clear that it’s still too hard to get started with Dgraph. Today, we’re excited to announce two upcoming changes to our offerings to make building with Dgraph the obvious choice.

The complete, fully open-source graph database

To-date, Dgraph has operated with two builds: one open source under the Apache License, Version 2.0 (APL) and the other with additional features under a commercial license. This is a historical approach for software-centric business models, holding back critical features that make it harder to get into production. We want Dgraph to be the standard for graph and AI knowledge stores (and Hypermode the best place to host it), so we’re changing this.

Starting with the Dgraph v25 release in early 2025, we will have a single open-source license (APL) and build for Dgraph.

We’re releasing these Dgraph Enterprise features in our Apache-licensed codebase in v25. Note that some current enterprise features will be refactored or retired as part of this transition.

• Binary Backups – increase resiliency of your cluster with full backups that feature faster restores than export/import operations
  
  

• Encryption at Rest – configure block-level encryption for advanced data security
  
  

• Audit Logging – record each database operation for deeper understanding and security analysis
  
  

• Multi-Namespace – share compute across schemas with data isolation and build multi-tenant apps

We’ll continue to provide our Dgraph Enterprise customers with a commercial offering that includes 24x7 author-backed support as well as advanced tooling for running Dgraph in complex, container-based environments.

Pay-for-use with Dgraph Serverless

Having the confidence of open-source portability is a key decision factor for many developers, but operationalizing a database to build a new feature isn’t a great starting point. Cloud-hosted offerings should be the fastest and most cost-effective path from a great idea to live backend ready to scale with your userbase.

The Dgraph Cloud entry tiers have allowed tens of thousands of developers to experiment and build prototypes before scaling their most promising ideas. Using shared compute, these on-ramps have suffered from noisy neighbors and low caps attempting to constrain the noise. Moving to dedicated compute with high availability is a significant step up in cost, making it difficult to grow together.

Dgraph already features a strong separation between compute and storage. Our team has been hard at work optimizing the cold-start time to unlock a new modality with a serverless expression of Dgraph. It will be initially optimized for development and prototypes, but we imagine this serving small-scale production apps in the future. This allows you to use compute in time-based slices when you have traffic rather than paying for idle capacity. The benefits are two-fold with a lower on-ramp and a smoother growth curve as your latest concept takes off.

The serverless offering has been built on rock-solid infrastructure and is available now as part of a private preview offering. If you’re interested in getting access, sign up here.

–

We’re also introducing a refreshed logo for Dgraph that you see at the top of this post. Graphs can be expansive, though much like a telescope, the power to spot and encode hidden structures is where they really shine! Want to try it on for size? We’ll be sharing new Dgraph SWAG first with our serverless preview participants.

With these changes, we’re further cementing Dgraph’s position as the most popular, open graph database. Whether you’re building with AI models today or setting the foundation for the future, Dgraph is the best place to start.

Alongside the launch of Modus, a framework for building code-first intelligent APIs, we’ve partnered with Hashnode for ModusHack. This global hackathon features a special cash prize for the best use of Modus with Dgraph.

What will you build?